The quest to save KeyForge, the first procedurally generated collectible card game

The story of KeyForge is a strange one. The collectible card game arrived with much fanfare in 2018 (including from us), boasting a procedural algorithm capable of generating some 32 billion different decks of cards all on its own. The allure was its surprise factor, since not even the game’s developers knew what was inside each box. Publisher Fantasy Flight Games quickly got a foothold in hobby stores and established a nascent organized play circuit; the game felt like it was poised to become the next big CCG. Then, in September 2021, the publisher announced it was no longer able to produce any more cards.

The messaging at the time was cryptic. Fantasy Flight simply said that the game’s sophisticated algorithm was “broken” and that it needed to be rebuilt “from the ground up.” That may certainly be true. But there was a much bigger problem, said Christian Petersen, the company’s co-founder, in a recent interview with Polygon: All of the software engineers that helped make the algorithm in the first place now worked for a different company.

Petersen founded Fantasy Flight in 1995. The Minnesota-based publisher earned a name for itself with Petersen’s own strategy game, Twilight Imperium, widely considered to be one of the largest and most complex board games ever made. That single super-popular game gave rise to one of the premiere tabletop publishing houses in the United States, responsible for KeyForge of course, but also for other games that were based on franchises like Star Wars and Lord of the Rings, and many more modern classics devised by its own creative teams.

In 2014 Asmodee, a large multinational corporation with dozens of popular board games under its umbrella, snatched up Fantasy Flight. Petersen left not long after that to set up a new business called Strange Stars. The engineers that could have helped rebuild KeyForge for Asmodee now worked for him. So, he did what any good businessman would do: He made an offer to buy back the rights to KeyForge.

“Asmodee delayed again for another six months or so,” Petersen said. “Maybe they didn’t like the amount of money I was willing to pay. Finally, they came back and we made the deal this June.”

Now Petersen, who has spent the last several years, among other things, developing software and manufacturing systems for the board game industry, is back in the publishing business. His first product is called KeyForge: Winds of Exchange, and as of publication it’s earned more than $1 million in crowdfunding on Gamefound.

Is that enough money to rebuild the algorithm and send the game back out into the wild? Only Petersen knows for sure. In any case, he told Polygon he firmly believes KeyForge is still worth saving. So too does the game’s new producer, Michael Hurley. Also a veteran of Fantasy Flight Games, he was among the executives in the room when co-creator Richard Garfield (Magic: The Gathering) first pitched a prototype using — what else? — a highly modified Microsoft Excel spreadsheet.

“There [were] a lot of macros in it,” Hurley said, drawing out the words to emphasize the size of the file involved. “It contained a list of every card name that [Garfield] had designed for the game. […] When he wanted to create a deck, he would just run some scripts, and it would basically generate a list of cards that were within the deck. [Then] he would pull the cards that the spreadsheet told them to pull, and then put it together.

“He would just do that repeatedly until he had […] a couple of dozen different prototype decks that he had generated in this way.”

But, in the original design, not all of the decks actually worked very well.

“He originally wanted all the decks to be completely random,” Petersen said, “so that you have no idea what you get. But we said, ‘No, that’s not going to work because there’s going to be such a variation in what you get that it’s going to be a problem for players.’”

What they ended up with was a much more structured system — recipe is probably the better word — for deck creation. The KeyForge algorithms, new and old, both work the same way. They first draw 12 existing cards from each of three houses, which are thematic factions that give texture to the in-game lore. The deck then gets a name and unique art on the back of each card, similarly generated by the same algorithm. But not every deck is created equally, and more powerful decks are adjusted (sort of like a handicap score in golf) for competitive play.

But every once in a great while the algorithm does something unusual, creating an ultra-rare card called a maverick. That’s a card originally designed to be part of one house, but switched to be part of another. Mavericks even get their own symbol printed on the border to call them out. If present in a unique deck, mavericks can become the core of powerful and unexpected strategies that can be difficult for other decks to counter.

Image: Ghost Galaxy

Sample cards from the proposed new run of KeyForge decks.

Petersen says that the project to rebuild the algorithm is coming along nicely, and it should be ready in time for the next batch of procedurally generated cards, mavericks and all, due out with KeyForge: Winds of Exchange in January 2023. Will it be soon enough to give the game a second chance at success? He remains hopeful, but pragmatic.

“The big question is, is the audience still around?” Petersen said. “It’s very difficult to relaunch an injured game. It’s almost impossible. I’ve had many times in my career where we [have said], ‘This game was injured. It’s hobbling along. It’s mostly dead. And we love the game, we think it’s really great. But what can we do?’ In most situations it’s just not worth it. […] Why try to revive it [when it] just doesn’t make economic sense?”

The crowdfunding campaign ends on Sept. 26. Expect pre-orders to open soon after it ends, however, and to continue for many months until release.